[Soot-list] A Strange problem about library-dependent program analysis in Soot.

[Sai Zhang]

Thanks Cheng.

I tried the -i command, and it seems still does not work.

I also try to use "--app" option as described in
http://www.sable.mcgill.ca/soot/tutorial/usage/.
This option intends to treat all class under my dir as application class,
and all jar file referenced
as lib classes.

However, a more strange thing happens: it only chooses to analysis all
classes under *junit *and
*org.junit* package (they are the dependent jar of the analyzed code).

Could anyone please give some more clues? Thanks!!!

-Sai


On Wed, Jan 27, 2010 at 11:35 AM, Cheng Zhang <cheng.zhang.stap at gmail.com>wrote:

> Hi, Sai,
>
> I once encountered a similar problem when I was using Soot to analyse a jar
> file (perhaps jTidy...).
> The problem seems to be that the classes in the jar file are considered as
> library classes (e.g. those in rt.jar),
> because some version of these classes have already been included in the
> runtime library.
>
> I am not so sure, but I guess the -i option may help.
>
> Best,
>
> Cheng
>
> On Wed, Jan 27, 2010 at 11:07 AM, Sai Zhang <racezhang at gmail.com> wrote:
>
>> Hi all:
>>
>> I am using Soot to do some data-flow analysis on several medium to large
>> scale programs, *which all use third-party ja*r.
>>
>> I am confused about Soot's output, I guess it should be a common problem
>> for newible. My code skeleton is shown as
>> follows (which  is enough to show  the problem):
>>
>> PS. I use *Java 1.6 with the latest Soot release (not eclipse plugin)*. I
>> set up the environment in eclipse galileo release.
>>
>> //ANALYSIS_DIR is the directory where all class files reside
>> String[] sootArgs = {"-cp", ".;" + ANALYSIS_DIR, "-pp",  "-process-dir",
>> ANALYSIS_DIR, "-f", "jimple",
>>
>> PackManager.v().getPack("jtp").add(
>>                 new Transform("jtp.myanalysis",
>>                         new BodyTransformer() {
>>                             protected void internalTransform(Body body,
>> String phase, Map options) {
>>                               * //implement my own analysis here*
>>                             }
>>                         }));
>>
>> //run soot
>> soot.Main.main(sootArgs);
>>
>> //print the analysis result, I omit the following code
>> System.out.print .....
>>
>>
>> The evaluated program depends on a set of  third-party jars,  so* I add
>> all jars  needed to the eclipse project references*, and use* -pp in
>> Soot's arg* to configure it.
>> That should be fine, right?
>>
>> However, when I run my program, Soot* neither issues a warning, nor throw
>> an exception, nor print any resul*t!
>>
>> The output looks like follows (Even the progress info: like Soot is
>> processing file XXX is missing):
>>
>> Soot started on Tue Jan 26 18:43:53 PST 2010
>> Soot finished on Tue Jan 26 18:43:56 PST 2010
>> Soot has run for 0 min. 2 sec.     //this processing time seems to be too
>> short
>>
>> Some known facts:
>>
>> 1. I feel really strange about this behavior (it *definitely should have
>> some result*, even if I change BodyTransformer() implementation to print
>> every method name it encounters).
>> 2. It is also unlikely that any dependent jars are missing (in that case,
>> soot will issue exceptions, right?)
>> 3. I doube-check that all classes are placed in the right dir, which
>> strictly follows the package name (if not, soot will throw exception)
>> 4. I also increase the analysis memory to -Xmx1024, so
>> OutOfMemoryException should not be the case
>> 5. I check there is no java file or other format files in the target
>> directory
>> 6*. what I feel most strange is that,* it works for some lib-dependent
>> program, but does not for other fews. In my case, it works on the whole
>> Daikon source tree (160KLOC), but fails on apache-common-collection
>> (80KLOC),
>>    so I guess that should be nothing to do with the program size.
>>
>>
>> I am wondering is that a common problem of daikon?  do I forget to
>> configure sth? or i use the wrong command/parameter? or could any one give
>> some hints? Any comments are highly appreciated!
>>
>> Thanks
>>
>>
>> --
>> Sai
>>
>> _______________________________________________
>> Soot-list mailing list
>> Soot-list at sable.mcgill.ca
>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>
>>
>


-- 
--
Sai Zhang
Computer Science & Engineering Department
University of Washington, Seattle, USA
Email: (saizhang at u.washington.edu)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.mcgill.ca/pipermail/soot-list/attachments/20100127/a35e4297/attachment-0001.html